I’ve encountered an unexpected behavior with the OpenAI API where the context from one session appears to influence another, despite using different API keys in separate applications. This cross-context issue seems to have emerged following the recent feature updates in the API.
- When using ChatGPT for certain discussions and then switching to another platform with a different API key, the AI in the second platform brings up topics or continues discussions from the ChatGPT session.
- In my chatbot tool, which uses its unique API key, I observed responses related to topics from a completely separate session in a VSCode AI assistant extension, also operating under a different API key.
This behavior suggests a lack of proper session isolation, leading to potential data privacy concerns and confusion in user interactions.
- Is this a known issue or a byproduct of the recent updates?
- What measures can be taken to ensure isolated contexts for different applications using distinct API keys from the same OpenAI account?
- Are there best practices to prevent such cross-context interactions?